Background: Lorlatinib is a potent, brain-penetrant ROS1/ALK tyrosine kinase inhibitor (TKI), which has demonstrated efficacy in advanced ROS1 fusion-positive (ROS1+) non-small cell lung cancer (NSCLC), including in patients (pts) previously treated with crizotinib. Despite initial benefit, however, most pts experience disease progression on lorlatinib. Mechanisms of resistance to lorlatinib in ROS1+ NSCLC are poorly understood. Methods: We analyzed repeat tumor biopsies derived from advanced ROS1+ lung cancer pts progressing on lorlatinib. Next-generation sequencing (NGS, n = 17) or whole exome sequencing (n = 1) was performed to detect mutations, indels, and copy number alterations. Results: Sixteen pts underwent a total of 18 repeat tumor biopsies after progression on lorlatinib. Fourteen had received prior crizotinib; two received prior crizotinib and entrectinib. Median duration of therapy on lorlatinib was 13.5 months (95% CI, 8.3-18.4). Among the 18 cases analyzed by sequencing, 7 (38.9%) harbored a ROS1 resistance mutation, including G2032R (4/18, 22.2%), S1986F/L2000V (1/18, 5.6%), L2086F (1/18, 5.6%), and G2032R/S1986F/L2086F (1/18, 5.6%). Of note, ROS1 L2086F was a novel resistance mutation not previously reported in the literature, but analogous to ALK L1256F (a lorlatinib-resistant ALK mutation). Structural modeling studies showed that ROS1 L2086F causes steric interference with binding of lorlatinib, as well as crizotinib and entrectinib. In addition to ROS1 kinase domain mutations, NGS analyses also identified MET copy number gain in a lorlatinib-resistant case, validated by fluorescence in situ hybridization as high-level focal MET amplification (MET/CEP7 copy number ratio 6.3) without a concomitant ROS1 resistance mutation. Duration of therapy on lorlatinib was significantly shorter in pts with a post-lorlatinib ROS1 resistance mutation compared to those without (8.3 vs 18.1 months; p = 0.005). Conclusions:ROS1 resistance mutations are observed in over one-third of cases progressing on lorlatinib, including the solvent front mutation G2032R and a novel L2086F mutation. These findings underscore the importance of developing next-generation ROS1 TKIs with activity against ROS1 mutations, and the need to elucidate ROS1-independent resistance mechanisms.